Assessing Allelic Dropout and Genotype Reliability Using Maximum Likelihood

A growing number of population genetic studies utilize nuclear DNA microsatellite data from museum specimens and noninvasive sources. Genotyping errors are elevated in these low quantity DNA sources, potentially compromising the power and accuracy of the data. The most conservative method for addressing this problem is effective, but requires extensive replication of individual genotypes. In search of a more efficient method, we developed a maximum-likelihood approach that minimizes errors by estimating genotype reliability and strategically directing replication at loci most likely to harbor errors. The model assumes that false and contaminant alleles can be removed from the dataset and that the allelic dropout rate is even across loci. Simulations demonstrate that the proposed method marks a vast improvement in efficiency while maintaining accuracy. When allelic dropout rates are low (0-30%), the reduction in the number of PCR replicates is typically 40-50%. The model is robust to moderate violations of the even dropout rate assumption. For datasets that contain false and contaminant alleles, a replication strategy is proposed. Our current model addresses only allelic dropout, the most prevalent source of genotyping error. However, the developed likelihood framework can incorporate additional error-generating processes as they become more clearly understood.